Has anyone seen the party breakdown for the Gallup sample?

[jefferson_dem]

You know, the one that has all the talking bobbleheads parroting about a chimpish-/-repuke comeback. In the past, Gallup has been plagued by wild swings in the PID of respondents, and they do not weight by PID. I suspect they oversampled Reps this time but have not seen any specifics. Also, note that they use *likely* voters this time as opposed to *registered* voters.

[tritsofme]

1. It would be highly irresponsible if they used "likely voters" for the

approval rating. Gallup has never done that in the past, and I doubt that's the case now, however with the Congressional races they made both registered and likely results available, which should happen this close to an election.

Party self ID in Gallup polls are inconsequential, they simply ask a respondent to identify with the Dems, GOP, or as an Indy at the end of the survey, there is no weighting, and if the sample is random, no bias.

[jefferson_dem]

2. Hmm. I was going on what they are reporting at pollster.com

The new survey also provides results to the generic Congressional ballot question among likely voters. While the Democrats lead among all registered voters (51% to 42%), the result is dead even (48%44% each) among those classified by Gallup as likely voters.

http://www.pollster.com/mystery_pollster/gallup_confirms_rise_in_bush_j.php

I think the "48%" is a typo and should read "44%".

As for weighting by PID, that's certainly a debatable methodological question. Some pollsters do weight, just as all weight for various other demographics such as gender, race, and age. Perhaps it depends on whether you believe PID is fairly static or more fluid. I recall Gallup's funky pre-election 2004 polls when they had 10 or more point swings in party loyalties from poll to poll. Do we really think that's the reality?

[tritsofme]

3. You can go to Gallup's website

and dig up their blog from about September or October 2004 where they go into great detail about their party ID methodology, it was a good resource and I probably posted it here dozens of times back then.

The swings seemed odd, but I really don't believe that party ID is as static as figures that can be compared to the census.